Parallel Preconditioning Technology In Electromagnetic Scattering From

Method of Moments combined with the fast multipole algorithm can accurately analyze the electromagnetic characteristics of the object.Due to the computing capacity limit of single computer, parallel computing is an essential solution to analyze large-scale object. The impedance matrix form Method of Moments is dense matrix when we want to solve the electromagnetic parameters, and the matrix produced by the large-scale or complex objects are usually poor conditions. When using an iterative solver for this equation, it usually appears slow convergence or even not convergence phenomenon. The preconditioning technique for the analysis of large-scale or complex objects is necessary.Sparse approximate inverse preconditioner can be used to solve large linear equations with slow convergence problem for its high parallel efficiency. Sparse approximate inverse requires a good sparse strategy and parallel strategy to achieve high efficiency. This article is based on FMM finest group, use the distance between basis functions and group center to determine whether the element should be selected to construct the approximate inverse matrix. Because the pecfect conductor and mixed dielectric/conductor object should be use different integral equations, the produced matrix equations are also different. Due to the dense mesh of the dielectric part, it will increase construction time, different sparse strategies are employed to the conductor and dielectric part respectively, can reduce construction time greatly. This paper also proposed several methods to improve the efficiency of the preconditioning, which are combined FGMRES with sparse approximation algorithm; current series expansion algorithm, multi-step spectral preconditioner based on sparse approximate inverse preconditioners. Numerical results shows the advantages of the above improved methods.When using the Method of Moments for the analysis of antennas, due to complex shape and tiny mesh of the source part, slow convergence phenomenon usually appears, therefore the parallel precondition technique for the analysis of antenna problems in this study is very important. This thesis focuses on the electromagnetic characteristics of the array antenna and reflective surface and satellite antenna, and gives the parametric analysis of the antenna. Numerical experiments show the preconditioning technique can greatly improve the efficiency in the analysis of these problems.